Archived Forum

Selecting the burden is simple: the ct acts as a current source, so the burden voltage is the secondary current multiplied by the burden resistance. The hard part is knowing whether the product of voltage and current is within the rating of the transformer, and few manufacturers state that directly.

Without downloading and studying the data sheets, the first is the only one that gives some indication of the VA rating. That one does indeed appear to be suitable, with a maximum of 20 A with the recommended 100 Ω burden, which at 16 A would give you 1.6 V rms (4.52 V p-p assuming a perfect sine wave) which allows 9.5% in hand. A 110 Ω burden would be OK to about 15.58 A, which is only just outside your specification (given worst-case components).

None are split-core (that is where the additional cost and loss of accuracy comes from) so you must be able to legally and safely disconnect the primary conductor to allow the ct to be fitted.

Disconnecting the primary conductor would not be a problem since this will be installed directly inside each domestic appliance.

You just have to be sure it is unplugged, cut the right conductor inside the appliance and install the CT.

So, any idea if this will have better resolution than the SCT's? This is the main concern, since I would like to measure low current, stand-by power too.

Thanks for your help Robert.

You need to do the sums. You can work out what voltage swing you will get at the ADC input for a given primary current, CT ratio and burden resistor, then compare that to the maximum possible input range, reduced a little to take account of component tolerances (these in effect reduce the range by a few percent - if your midpoint bias is wrong by 1%, the burden is wrong by 1% and the ct ratio wrong by 2%, you have lost about 4% of the available range), therefore you need to reduce your burden value to guarantee to avoid clipping and so you lose range.

Resolution isn't a property of the CT, it is down to the ADC, and quite often also down to how well you can filter and screen the analogue input from the noise emanating from the digital circuitry.

If you look at Building Blocks, the CT and burden calculations are there, also there is something about ADC resolution. I did have a spreadsheet somewhere, where I investigated the effect of noise and offset on the accuracy, and Trystan made it into a web page (since removed). Experience says you can have reasonable accuracy with an emonTx down to about 1% of the maximum current, below that the errors increase, depending significantly on where the quiescent input point sits in relation to the ADC steps. This curve (calculated values only, neglecting noise) shows how the ratio converted rms / true rms (i.e. the conversion error) varies with the offset over a range of input amplitudes. You cannot know what offset you will have in practice - it's essentially random and variable.